Recommended prerequisite for participation in the module

Content, progress and pedagogy of the module

Cameras capture visual data from the surrounding world. Building systems which can automatically process such data requires computer vision methods. Students who complete the module will understand the nature of digital images and video and have an inside into relevant theories and methods within computer vision and an understanding of their applicability.

Learning objectives

Knowledge

• Must have knowledge about the primary parameters of a camera system
• Must have knowledge about the representation and compression of digital images and video signal
• Must be able to understand the general framework of image processing as well as the basic point and neighborhood operations, i.e., binarization, color processing, BLOB analysis and filtering
• Must be able to explain the principles behind invariant feature point descriptors such as SIFT and Harris corners.
• Must have knowledge of different motion analysis methods, such as background subtraction and optical flow
• Must be able to understand the tracking frameworks such as the Kalman filter, mean-shift and the particle filter
• Must be able to understand different shape analysis methods such as active-shape models, procrustes, Hungarian method

Skills

• Must be able to apply stereo vision to generate 3D date from two or more cameras. This implies projective geometry, camera calibration, epipolar geometry, correspondence and triangulation
• Must be able to apply advanced 2D segmentation methods such as Hough transform, compound morphology, and histogram-of-oriented histograms.
• Must be able to demonstrate understanding of error propagation techniques as a tool for performance characterization of computer vision based solutions

Competences

• Must be able to learn further computer vision methods and theories, and select an appropriate solution for a given problem

Type of instruction

As described in § 17.

Exam

Exams